Krah et al. BMC Evolutionary Biology (2018) 18:119 https://doi.org/10.1186/s12862-018-1229-7

RESEARCHARTICLE Open Access Evolutionary dynamics of host specialization in wood-decay fungi Franz-Sebastian Krah1,2* , Claus Bässler2, Christoph Heibl2, John Soghigian3, Hanno Schaefer1 and David S. Hibbett4

Abstract Background: The majority of wood decomposing fungi are mushroom-forming Agaricomycetes, which exhibit two main modes of plant cell wall decomposition: white rot, in which all plant cell wall components are degraded, including lignin, and brown rot, in which lignin is modified but not appreciably removed. Previous studies suggested that brown rot fungi tend to be specialists of gymnosperm hosts and that brown rot promotes gymnosperm specialization. However, these hypotheses were based on analyses of limited datasets of Agaricomycetes. Overcoming this limitation, we used a phylogeny with 1157 species integrating available sequences, assembled decay mode characters from the literature, and coded host specialization using the newly developed R package, rusda. Results: We found that most brown rot fungi are generalists or gymnosperm specialists, whereas most white rot fungi are angiosperm specialists. A six-state model of the evolution of host specialization revealed high transition rates between generalism and specialization in both decay modes. However, while white rot lineages switched most frequently to angiosperm specialists, brown rot lineages switched most frequently to generalism. A time- calibrated phylogeny revealed that Agaricomycetes is older than the flowering plants but many of the large clades originated after the diversification of the angiosperms in the Cretaceous. Conclusions: Our results challenge the current view that brown rot fungi are primarily gymnosperm specialists and reveal intensive white rot specialization to angiosperm hosts. We thus suggest that brown rot associated convergent loss of lignocellulose degrading enzymes was correlated with host generalism, rather than gymnosperm specialism. A likelihood model of host specialization evolution together with a time-calibrated phylogeny further suggests that the rise of the angiosperms opened a new mega-niche for wood-decay fungi, which was exploited particularly well by white rot lineages. Keywords: Wood-decay fungi, Decay mode, White rot, Brown rot, Host specialization, R package rusda

Background chains of glucose subunits that can take on a recalcitrant About 2000 billion tons of carbon is present in terres- crystalline form [6]. Hemicelluloses are matrix polysac- trial ecosystems [1], of which 550 billion tons are fixed charides consisting of various heteropolymers, e.g., of in vegetation [2]. In forest ecosystems, most plant bio- xylans and glucomannans [7]. Lignin is a complex aro- mass is stored in the form of dead wood [3]. Woody matic polymer that is resistant to hydrolytic degradation plant cell walls consist mainly of the lignocellulose com- [8]. The amount of cellulose in woody plants is 40–50% plex which is composed of the polymeric polysaccha- of the wood dry weight and for hemicelluloses and lignin rides cellulose, hemicellulose and lignin heteropolymers 15–30% each. The plant biomass further consists of [4, 5]. Cellulose is a macropolymer consisting of linear macromolecules such as lipids, waxes, proteins and phenolic compounds [3]. The most efficient agents of * Correspondence: [email protected] the decay of the lignocellulose complex are saprotrophic 1Plant Biodiversity Research Group, Center for Food and Life Sciences fungi, which therefore play pivotal roles in the cycling of Weihenstephan, Technische Universität München, Freising, Germany 2Baverian Forest National Park, Grafenau, Germany carbon [9] and nutrients [10] in the forest ecosystem. Full list of author information is available at the end of the article Wood is produced by angiosperms and gymnosperms,

© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 2 of 13

which together comprise more than 60,000 species [11]. are distinguished by high copy numbers of genes encod- Angiosperms regularly have lower amounts of lignin ing different carbohydrate-active enzymes (CAZymes) than gymnosperms, whereas angiosperms regularly have which are classified based on the CAZy database [27]. In higher amounts of cellulose than gymnosperms [12, 13]. general, CAZymes, which act on crystalline cellulose are Further, angiosperms often have lower amounts of more abundant in white rot genomes compared with non-structural secondary compounds (plant extractives) brown rot [24]. Glycoside hydrolase (GH) families (e.g., than gymnosperms [12–15], with some exceptions, e.g., GH6 and GH7, including cellobiohydrolases) are more species of the genera Quercus, Fagus or Malus [16]. abundant in white rot compared with brown rot fungi The main agents of wood decay are members of the [24]. Further, lytic polysaccharide monooxygenases class Agaricomycetes (Basidiomycota). Agaricomycetes (LPMOs) from the AA9 family are more abundant in contains about 21,000 species with a worldwide distribu- white than brown rot fungi [24]. Apart from Agaricomy- tion, including many lifestyles, e.g. mycorrhizal symbi- cetes, LPMOs can be found in Ascomycetes and Mucor- onts, pathogens, and saprotrophs. Most saprotroph omycotina [21, 24]. Finally, lignin-degrading class II fungi within the Agaricomycetes are dead wood decaying peroxidases (AA2) and other heme-containing peroxi- fungi. Dead wood decay modes can be classified as ei- dases are more common in white rot, and reduced or ther white or brown rot. Brown rot fungi attack cellulose absent in brown rot fungi [26] (for mechanisms of action but do not significantly degrade lignin [17], resulting in see Hofrichter et al. [28]). The most recent common an- a brownish residue that breaks into cubical fragments, cestor of Agaricomycetes was a white rot species (based whereas white rot fungi degrade both cellulose and lig- on an inferred expansions of AA2 and other lignocellu- nin [18], leaving a bleached fibrous residue (Fig. 1). lolytic enzymes) with at least four independent origins of Hemicellulose can be degraded by both brown and white brown rot, correlated with parallel losses of genes en- rot fungi [19]. Whereas dead wood is mainly decayed by coding diverse CAZys, and the complete loss of lignino- Agaricomycetes, in plant litter decay, Ascomycota play a lytic class II peroxidases (AA2) [24], making reversals to significant role along with litter-decomposing Agarico- white rot unlikely. This white rot ancestor likely lived mycetes [19, 20]. Other decay modes are also present, roughly 290 (+/− ca. 70) million years ago (MYA) [24]. such as “soft rot” in some Ascomycota and “grey rot” in Analyses of a sample of 62 genomes by Nagy et al. [29] some Basidiomycota such as Schizophyllum commune suggested that expansions of cellobiohydrolases (GH6, [21]. Although many ectomycorrhizal fungi are partially GH7), LPMOs (AA9), and other plant cell wall degrad- saprotrophic, their decay abilities are considered mar- ing enzymes occurred early in the evolution of Agarico- ginal compared to wood decay fungi [22, 23]. mycetes, prior to the expansion of class II peroxidases The enzymatic basis of the differences between white (AA2). rot and brown rot has been studied extensively in com- Gilbertson [30] investigated ecological differences be- parative genomic analyses [21, 24–26]. White rot fungi tween white and brown rot decay modes, noting that

a c

b d

Fig. 1 Brown and white rot residues and fungal fruit bodies. a) Brown rot residue, b) brown rot fungus, Fomitopsis pinicola (Polyporales, Fomitopsidaceae), c) white rot residue, d) white rot fungus, Fomes fomentarius (Polyporales, Polyporaceae). Photos by F.-S. Krah (a,b,c) and Heinrich Holzer (d) Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 3 of 13

brown rot fungi preferentially occur on gymnosperm interpretations, we re-sampled a single species per genus hosts [31]. Gilbertson thus suggested a correlated evolu- from our final dataset (hereafter “one-genus-subset”)and tion of brown rot decay mode and gymnosperm repeated the analyses described below 100 times. specialization. Hibbett and Donoghue [32] tested Gil- To gather data on host associations, we used the R bertson’s hypothesis using phylogenetic comparative package “rusda”, written for this study, as an interface to methods. Their results suggested that the evolution of the USDA Fungus-Host Distribution Database (FHDD) brown rot was correlated with the evolution of exclusive [33]. The FHDD contains fungus-host combinations, but decay of gymnosperm hosts. However, this inference was does not provide information on the occurrence fre- made from a dataset with limited taxonomic sampling, quencies on a particular host (other than the number of with only a total of 130 species [32]. published records on each host). The “rusda” package To assess the evolution of decay modes and patterns makes it possible to retrieve (“query”) host data for fun- of host specialization among wood decay fungi in Agari- gal species, and vice versa. For a detailed description, comycetes, we utilized a time-calibrated mega-phylogeny basic usage and evaluation of the R package “rusda”, see approach and drew on the extensive Fungus-Host Distri- Additional file 1 and for repositories refer to Availability bution Database built by the United States Department of data and material. of Agriculture (USDA) [33]. We then used this To retrieve host associations from the FHDD we used mega-phylogeny and host associations, which encom- the function associations, which takes an input of species passed 1157 species from 14 orders, to test two hypoth- names and provides an output list of fungus-host combi- eses: (1) brown rot fungi occur primarily on nations. As input we used the NCBI taxonomy for gymnosperm hosts; and (2) brown rot fungi switched fungi and re-classified the order level where necessary more frequently towards gymnosperm hosts than white (Additional file 1: Table S3). We then produced a rot lineages. We further use this large-scale dataset to dataset of plant phyla by matching host genera to the investigate white rot specialization pattern and mecha- Spermatophyta taxonomy downloaded from NCBI nisms, a topic currently neglected due to a focus on taxonomy using the R package “megaptera” [42]. Thus, we specialization pattern of brown rot fungi. retrieved the phylum information “Acrogymnosperma” and “Magnoliophyta” for each host species. We refer to Methods “Magnoliophyta” as angiosperm (A) and “Acrogymnos- Trait data and character matrix perma” as gymnosperm (G) and stored the number of To test our hypotheses, we gathered data on decay mode gymnosperm and angiosperm associations for each fungus and host associations for Agaricomycetes. For decay species in a table. Species which did not belong to either mode, we used the “decay.type” as published in Tedersoo Acrogymnosperma or Magnoliophyta were deleted from et al. [34], which is available on the genus level, and we the dataset. We further deleted all non-woody plants also conducted a literature search for additional genera. based on the woodiness dataset which classified more Tedersoo et al. (2014) investigated lifestyle-dependent glo- than 35,000 plants into woody and non-woody [43]. Thus, bal fungal diversity and therefore coded the trophic status the final host dataset included only woody plants from (six states, e.g., biotroph), the lifestyle (17 states, e.g., ecto- Acrogymnosperma or Magnoliophyta; seedless vascular mycorrhizal) and decay type (four states, e.g. white rot, plants, bryophytes, algae, and non-plant hosts were brown rot) for more than 10,000 genera. We used only excluded. The FHDD covers mainly temperate North species with either white or brown rot in our analysis and America and Europe [33]. excluded other lifestyles (e.g., mycorrhizal). This gave us The host association data were used to calculate the the decay mode of particular species in the genera. We number of angiosperm and gymnosperm host species then extrapolated this decay mode to the remaining spe- for each fungus species. We defined the “gymnosperm cies of a genus (with one exception, see below). Our justi- association” by dividing the number of gymnosperm fication is that decay mode has often been a focus of host tree species (NG) by the sum of the number of taxonomists and thus was widely used to distinguish gen- angiosperm (NA) and gymnosperm host tree species: era such as Antrodia and Antrodiella [35],Lentinusand gymnosperm associations [%] = NG/(NG +NA). Thus, a Neolentinus-Heliocybe [36], and Daedalea and Daedaleop- gymnosperm association of 100% means that a fungus is sis [37]. We found only three genera where more than one reported exclusively on gymnosperm hosts in the decay mode has been reported: Clitocybula [38, 39], Fungus-Host database, whereas 0% means only angio- Hyphoderma [40], and Mucronella [40, 41]. Clitocybula sperm hosts are reported. We classified host preferences and Mucronella were deleted from the dataset because no into three states: (1) generalism, (2) angiosperm host data were available. For Hyphoderma we used only specialization, or (3) gymnosperm specialization. Based the two species where decay mode references were found. on the distribution of gymnosperm association [%] To estimate how this strategy might affect our (Fig. 4c), we defined specialization based on the Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 4 of 13

gymnosperm association [%] with a threshold of ≥90% (Additional file 1: Table S4). We used the R package for gymnosperm specialization and a threshold of ≤10% for “megaptera” to calculate the identity (proportion of nu- angiosperm specialization (hereafter “90–10 specialization”). cleotides identical) and coverage (proportion of nucleo- Previous studies used exclusivity as a measure of host asso- tide positions in common) with the reference. Based on ciation [32], but missing or incorrect data for a single fungus the coverage and identity values, thresholds can be ad- observation may then lead to misclassification of a species. justed aiming to maximize both quality and number of Nonetheless, we also inferred our final model (see Statistics taxa. The default values are 0.75 for identity and 0.5 for and models of host specialization) using the exclusivity cod- coverage. Based on visual inspection of the alignments, ing (hereafter “100–0exclusivity”). However, in the exclusiv- we chose identity thresholds between 0.5 and 0.75 and ity coding, generalists and non-exclusive specialists are coverage thresholds between 0.25 and 0.5 for the seven coded in one state (“generalists”) and thus results might be gene regions. All sequences outside these limits were hard to interpret. discarded. We aligned the remaining sequences for each gene re- Mega-phylogeny approach gion separately, using GUIDANCE2 [48, 49] with the To test dynamics of host switching, we used phylo- multiple sequence alignment program MAFFT [50]. genetic comparative methods (PCMs). For this pur- GUIDANCE2 computes a reliability score for each col- pose, we applied a mega-phylogeny approach using umn based on alternative alignments produced by boot- the R package “megaptera”, a pipeline for large-scale strap guide trees and four co-optimal alignments based automated sequence-retrieval and alignment [44](ver- on each bootstrap alignment, created by the heads or sion available on https://github.com/heibl/megaptera). tails algorithm [51]. We passed the resulting column The mega-phylogeny approach aims at maximising taxon score as character weights to the phylogeny inference sampling integrating previous knowledge (e.g. taxonomic program RAxML (flag -a; see additional details on information, backbone trees) into the tree inference [45]. phylogenetic inference below) rather than filtering the For our mega-phylogeny approach, we used a backbone alignment using the column score, which is not recom- guide tree based on phylogenomic analyses [24, 26, 29]to mended [52]. We used IQ-TREE version 1.5.3 with spe- provide information for deep splits (order level), as resolv- cification “-TESTMERGEONLY” [53, 54] to select a ing such ancient divergences can be difficult due to partition scheme among the gene regions. IQ-TREE sequence saturation [45]. Further, mega-phylogeny ap- found six blocks as the best partitioning scheme (mer- proaches often lead to a high number of gaps or missing ging the 5.8S rRNA and 28S rRNA into one partition; data, often more than 90% (e.g. Smith et al. [45]). To re- Additional file 1: Table S1). The final alignment had duce the bias of missing data, we computed a reliability 37,466 sites and the proportion of gaps was 92.07% with measure for each column of the alignment, which is then 16,814 distinct alignment patterns. supplied to the tree inference program. In this way, uncer- We produced a comprehensive backbone guide tree by tain regions in the alignment are down-weighted in the first assembling an order-level “genomic” based back- phylogeny inference step. bone tree (Additional file 1: Figure S1 A) from the litera- First, we used the R package “megaptera” to download ture [21, 26, 29] and then attaching all species on the all sequences for the species with decay mode and host order-level tips of the genomic backbone tree (Add- association information from GenBank [46] (queried itional file 1: Figure S1 B). We performed maximum February 2017). We selected seven DNA regions: 18S, likelihood estimation, using the concatenated superma- 28S and 5.8S rRNA (nuclear ribosomal RNA genes), trix of the seven DNA regions, with RAxML [55] on the genes encoding RNA polymerase b (rpb1, rpb2), transla- CIPRES Science Gateway v.3.3 (RAxML -HPC2 on tion elongation factor 1 (tef1), and ATP synthetase XSEDE 8.1.11) [56, 57] under the GTRGAMMA model (atp6). We chose the rRNA regions to obtain high spe- with partitioning as described above, the GUIDANCE2 cies numbers and the other regions for resolution of column score (flag –a) and the comprehensive backbone deeper nodes [47]. Only sequences of samples identified tree (flag –g). We subsequently conducted 1000 approxi- to species level were accepted. mate Shimodaira–Hasegawa likelihood ratio tests We used single sequences where only one sequence (SH-aLRT branch support). SH-aLRT which are fast, ac- for a particular species and DNA region was available. If curate and robust even for larger phylogenies [58]. multiple sequences were available, all sequences of the We estimated divergence times of the resulting phyl- same DNA region and organism (putatively conspecific ogeny using penalized likelihood as implemented in the sequences) were aligned and a majority rule consensus R function chronos from the R package “ape” [59]. We sequence was calculated. In the next step, all sequences used two calibration points, a Late Cretaceous mush- were compared to three to six Agaricomycotina refer- room fossil Archaeomarasmius legetti [60], which bears ence sequences for each DNA region as a quality check a strong resemblance to extant Agaricales (particularly Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 5 of 13

Marasmiaceae), and a Middle Eocene ectomycorrhizal multi-state version of a continuous-time Markov model, fossil, which has been interpreted as a representative of where the Markov process is characterized by a Q-matrix. Boletales [61]. We followed the strategy of Kohler et al. The Q matrix specifies the transition rates between the [26] and used the ectomycorrhizal fossil to calibrate character states and hence the model of discrete character Boletales with a stem age of 40–60 MYA and A. legetti evolution. All models were based on our six-state character to date Agaricales with a stem age range of 70–110 coding and the transition rate matrix was a 6 × 6 matrix: (1) MYA. We also tried the approach of [24] and used 50 white rot/angiosperm specialist, (2) brown rot/angiosperm and 90–94 MYA as age priors, which yielded almost specialist, (3) white rot/gymnosperm specialist, (4) brown identical divergence time estimates (results not shown). rot/ gymnosperm specialist, (5) white rot/generalist, and (6) We applied chronos with three different models of substi- brown rot/generalist. tution rate variation among branches: “relaxed”, “corre- The first model allows for all transitions to occur in lated” and “strict” andcomparedthemodelfitsusingɸIC single steps, e.g. an angiosperm specialist can switch dir- [62]. The “correlated” model had lowest ɸIC values and ectly to a gymnosperm specialist without first passing thus was used for further analysis. We are aware that penal- through a generalist state. Further, in this model transi- ized likelihood does not make use of the sequence data and tions between white rot and brown rot are allowed in does not incorporate phylogenetic uncertainty. However, both directions. All models allow white rot to brown rot algorithms that perform joint inferences of the tree and transitions. We call this the “Uncorrelated” model, be- divergence times currently do not implement an option for cause switches between the states are not conditioned character weights, e.g. BEAST [63] or character weights on previous states. This model may not be biologically and guide tree, e.g. ExaBayes [64]. realistic. Transitions from an angiosperm specialist to a To account for phylogenetic uncertainty at nodes with gymnosperm specialist may require a transition first low support values, we produced alternative trees based through a generalist, before passing to a gymnosperm on the maximum likelihood phylogeny (Additional file 1: specialist, and thus could require two “steps”. Thus, we Figure S2). We created hard polytomies on nodes with coded further models implementing correlated SH-like support values < 80 based on the non-ultrametric (dependent) character evolution. In the second model, ML tree (Additional file 1:FigureS3).Wethenusedthe we prohibited transitions leading directly from one spe- function multi2di from the R package “ape” [59]and cialist to another by setting the direct transition parame- resolved the polytomies randomly and used chronos (as ters to zero. We call this the “Correlated hosts” model. described above) to estimate divergence times. We re- Both the “Uncorrelated” and the “Correlated hosts” peated this 100 times and summarized the dated trees model allow for brown rot to white rot reversals. How- using TreeAnnotator [65] to calculate a maximum clade ever, brown rot evolution is correlated with complete credibility tree (MCCT) with the node option “Common losses of genes encoding ligninolytic class II peroxidases ancestor heights” (because the nodes did not share the (AA2) and reductions in other decay enzymes, making same ancestors since polytomies were created at random). reversals to white rot unlikely [24]. Accordingly, we con- We displayed confidence intervals of the divergence time structed a third model where we further disallowed tran- estimates as HPD (highest posterior density) for the sitions from brown rot states to white rot states. We call brown rot clades and the root. Furthermore, we use the this the “Correlated hosts – norev” model. For the cod- 100 ultrametric trees as input for the transition rates esti- ing of the Q matrices, see Additional file 1: Figure S4. mation to measure robustness of the results against phylo- We fitted the three models with equal rates (ER) and genetic uncertainty. all rates different (ARD) and compared the fit of the models by Akaike’s information criterion (AIC) [69] Statistics and models of host specialization from the log-likelihoods. For model selection we applied We first tested preferences of host species among extant a simple root state with equal weights among the six fungi of the two decay modes using a phylogenetic linear character states (root.p = NULL). Brown rot has been model in the R package “phylolm” [66]. We tested whether shown to evolve repeatedly from white rot ancestors [70, the number of host species (host range) differed between 71], so we applied an additional root state treatment decay modes as a binary predictor variable. As an evolu- which only allows white rot as root state. Thus, after tionary model for the residual variance-covariance matrix model selection we ran the final (best) model using an we used the lambda model [67]. The number of host tree additional root state coding, which assumed zero prob- species was log10-transformed. ability for brown rot and equal probabilities for each of We modelled dynamics and pattern of host specialization the three white rot states, and compared the models. evolution in white and brown rot lineages using multistate Another framework to estimate pattern of host evolu- likelihood-based models. We used the function rayDISC tion is the coding as three independent binary states: from the R package “corHMM” [68], which implements a white rot – brown rot; angiosperm – no angiosperm; Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 6 of 13

gymnosperm – no gymnosperm (e.g. using the function Results corDISC, from the R package “corHMM”). However, this Our core dataset consisted of 1157 fungal species, in- model requires unobserved states (no angiosperm and cluding 126 brown rot and 1031 white rot species. Based no gymnosperm host). Such unobserved states may yield on the 90–10 specialization coding, we found 205 gym- high rates as a methodological artefact [72]. Thus, we nosperms specialists, 565 angiosperm specialists and 387 decided to use the multi-state implementation in the generalists (for tip state frequencies, see Additional file 1: function rayDISC. Table S2). Our time-calibrated phylogeny contains five brown rot Phylogenetic signal clades (Fig. 2, all clades had SH-like support values We computed phylogenetic signal in decay mode, gymno- above 90, Additional file 1: Figure S3), including two in sperm association, and the six-state character coding (as Polyporales, one in Gloeophyllales, one in Agaricales defined above). For the decay mode (binary state) we used and one in Boletales. Clade 1, the Auriporia-Crusto- the phylogenetic D statistic, which is calculated as the derma clade, within the Polyporales includes Laetipora- sum of sister-clade differences based on reconstructed ceae, Sparassidaceae, Dacryobolaceae pro parte values on all nodes of the tree [73]. The observed D is (Dacryobolus karstenii), Crustoderma and Pycnoporellus. then compared against (1) a random expectation (random Clade 2, the Antrodia-Fomitopsis clade, within the Poly- shuffling of trait values along the tips), and (2) a trait sim- porales includes Fomitopsidaceae, Dacrybolaceae pro ulated according to a Brownian motion model of character parte (Spongiporus, Oligoporus, Postia pro parte) and evolution along the tree, after the values were converted Fibroporia gossypinum. Clade 3, the Gloeophyllum-Neo- to a binary according to a threshold. For the computation lentinus clade, falls within the Gloeophyllales. Clade 4, we used the function phylo.d in the R package “caper” [74] the Fistulina clade, falls within the Agaricales (Fistulina with 1000 permutations. pallida and F. antarctica). Clade 5, the Serpula-Hygro- For the gymnosperm association we calculated two phoropsis clade, falls within the Boletales (Fig. 2). measures of phylogenetic signal: Pagel’s lambda [67] using the function phylosig from the R package “phy- Phylogenetic signal tools” [75], and phylogenetic correlograms using the For decay mode, we found a phylogenetic D value of − function phyloCorrelogram from the R package “phylo- 0.38, which had a high probability resulting from a singal” [76]. Lambda measures the phylogenetic depend- Brownian motion phylogenetic structure (P = 0.998) and ence of a trait under the assumption of a pure Brownian a corresponding low probability resulting from a random motion model of evolution. Lambda is a transformation phylogenetic structure (P = 0.00, Fig. 3a). We found a (weight) of the variance-covariance matrix, if other fac- lambda value of 0.73 for gymnosperm association and an tors than the phylogenetic history had an effect on the increasing phylogenetic signal towards the tips (Fig. 3b, trait. If lambda equals 1 the model fits a Brownian mo- red and blue lines indicates significance). For the tion model of evolution. Phylogenetic correlograms six-state character coding, we found that the observed measure phylogenetic signal in dependence of the phylo- parsimony score was significantly smaller than under a genetic distance (that is distance in branch lengths). For random expectation (Fig. 3c). a single trait, phylogenetic signal is measured as the autocorrelation (Moran’s I) based on a sequence of Host preferences among decay fungi phylogenetic weights matrices differing in their mean We assessed host preferences among extant decay fungi (phylogenetic distance if method = “lag-norm”). We con- based on the average number of host tree species. White ducted 100 bootstraps for 100 points to generate a confi- and brown rot fungi did not significantly differ in their dence interval. If the confidence interval falls below or average number of host tree species (phylogenetic re- above 0 the signal becomes significant. We rescaled the gression, Fig. 4a, b, statistics Additional file 1: Table S5), phylogeny to a tree height of 1 for this analysis. although visible trends suggested that white rot species For the six state character coding we calculated the have a larger average host range on angiosperms (Fig. 4a), phylogenetic signal following the method described in while brown rot species have a larger average host range Bush et al. [77] (function phylo.signal.disc, the script is on gymnosperms (Fig. 4b). The histogram of the gymno- available at: https://github.com/juliema/publications/ sperm association showed a bimodal distribution with blob/master/BrueeliaMS/Maddison.Slatkin.R). A parsi- two peaks towards the ends of the distribution, repre- mony score of the discrete trait along the tree is com- senting extremes of angiosperm vs. gymnosperm pared to a randomized parsimony score inferred by specialization (Fig. 4c). Thus, among the specialized randomizing tip states. If the parsimony score falls out- decay fungi most occur exclusively on either angiosperm side the random distribution, this indicates a higher con- or gymnosperm hosts (Fig. 4c). Based on the gymno- servation than under a random expectation. sperm association we found that 51% of white rot Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 7 of 13

appearance of Brown rot clades: tricolpate pollen 1 Auriporia-Crustoderma 2 Antrodia-Fomitopsis 1 Polyporales 2 3 Gloeophyllum-Neolentinus 4 Fistulina 5 Serpula-Hygrophoropsis Russulales fossil Gloeophyllales 3 Corticiales

estimated angiosperm (crown) origin Agaricales 4 estimated Boletales gymnosperm 5 Atheliales (crown) origin Hymenochaetales Trechisporales Gomphales Auriculariales Sebacinales Paleo- Neo- Carboniferous Permian Triassic Jurrasic Cretaceous gene gene Cantharellales 300 250 2000 51 100 50 0 Million years Fig. 2 Agaricomycetes species-level dated tree of wood-decay fungi. The species-level chronogram is based on the maximum clade credibility tree (MCCT) topology, dated with two fossils indicated by black points. Clades: (1) Auriporia-Crustoderma (Polyporales), (2) Antrodia-Fomitopsis (Polyporales), (3) Gloeophyllum-Neolentinus (Gloeophyllales), (4) Fistulina (Agaricales), (5) Serpula-Hygrophoropsis (Boletales). Blue bars on the brown rot crown nodes indicate confidence intervals (HPD) from 100 alternative trees. For a detailed maximum likelihood phylogeny with SH support values and tip labels, see Additional file 1: Figure S3 species are specialized to angiosperm hosts, whereas Of the five brown rot clades (Fig. 2), two consisted of 27% of brown rot fungi are specialized on angio- mainly generalist species (Polyporales clades: Auriporia- sperms (Fig. 4d). Among brown rot fungi, however, Crustoderma and Antrodia-Fomitopsis). Two clades consist we found a higher proportion of generalists and of mainly gymnosperm specialists (Gloeophyllales: gymnosperm specialists than in white rot fungi Gloeophyllum-Neolentinus; Boletales: Serpula-Hygrophor- (Fig. 4d). opsis). One clade consists of mainly angiosperm specialists

a Decay mode b Host association [%] c Host association - 6-states 10 0.15 200 90-10 specialization 8 observed lambda = 0.73 150 exclusivity 6 0.10 100 random 4 0.05 Density 2 Frequency 50 observed

Phylogenetic singal 0.00 0 0 0.0 0.5 1.0 0200400600 400 450 500 Phylogenetic signal D Phylogenetic distance Parsimony score Fig. 3 Phylogenetic signal for decay mode, and two measures of gymnosperm association. a) Phylogenetic signal D for decay mode (binary variable). A value smaller than 0 indicates strong conservatism. b) Pagel’s lambda and phylogenetic correlogram for gymnosperm association. A lambda value of 0.73 indicates non-random trait evolution which is not as conserved as Brownian motion. The phylogenetic signal increased towards the tips. Displayed is the mean phylogenetic signal with a 95% confidence interval resulting from 100 bootstraps. c) Phylogenetic signal C for the six-state coding. The observed value outside of the random expectation distribution indicates conservatism Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 8 of 13

a n.s. b n.s. c angiosperm 315 99.0 500 specialism 99 gymnosperm 30.6 400 specialism 31 300 9.0 9 200 2.2 Frequency 2 100 generalism Number of hosts 0 0.0 0 Brown White Brown White 0 20406080100 rot rot rot rot Gymnosperm association % d e 100 25 80 20 60 15 40 10 20 host species [%]

Scaled number of 5 Number of species Auriporia-Crustoderma Antrodia-Fomitopsis Gloeophyllum- Neolentinus Fistulina Serpula- Hygrophoropsis 0 1 2 3 4 5 Brown White 0 rot rot Brown rot clades f 100 500 400 80 300 60 200 40

100 host species [%] 20 Scaled number of Number of species

0 0 White rot

Agaricales Athelialesiculariales Corticiales GomphalesPolyporalesRussulalesSebacinales Aur Cantharellales Trechisporales Amylocorticiales Hymenochaetales White rot by orders Fig. 4 Distribution of gymnosperm association among wood-decay fungi and five major brown rot lineages within Agaricomycetes. a) Number of angiosperm host tree species for white and brown rot species. b) Number of gymnosperm host tree species for white and brown rot species. Note the log scale of the y-axis and that the values were back-transformed. Significances were inferred using phylogenetic regression (Additional file 1: Table S5). c) Bimodal distribution of gymnosperm association of wood-decay fungi. The six-state character coding was based on the gymnosperm association. A gymnosperm association above 90% was classified as gymnosperm specialist, below 10% as angiosperm specialist and others as generalists (“90–10 specialization”, for details see Trait data and character matrix). d) Scaled number of host tree species grouped by the 90–10 specialization coding. e) Number of species for the five observed brown rot clades (Fig. 2). f) Number of species for white rot fungi and scaled number of host tree species grouped by the 90–10 specialization coding. Note that Amylocorticiales, Gomphales and Sebacinales had less than five host data points. Angiosperm tree image by Michele M. Tobias (see Acknowledgements)

(Agaricales: Fistulina)(Fig.4e). The two Polyporales thus we did not interpret host associations for them. clades, Auriporia-Crustoderma and Antrodia-Fomitopsis, White rot species within six orders were primarily however, also display a considerable amount of angiosperm angiosperm specialists (Agaricales, Auriculariales, Cor- specialists, exceeding gymnosperm specialists (Fig. 4e). ticiales, Hymenochaetales, Polyporales, Russulales) Twelve of the 14 orders in our dataset contained white (Fig. 4f). White rot species within three orders were pri- rot lineages (Fig. 4f). Three of these had less than five marily generalists (Atheliales, Cantharellales, Trechispor- species (Amylocorticiales, Gomphales, Sebacinales) and ales) (Fig. 4f). Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 9 of 13

Dynamics of host switches Based on the models of discrete trait evolution describ- ing host switching dynamics among white and brown rot fungi, we found the “Correlated hosts – norev” 0.001 Brown rot model, with all rates different, as the best model (model White rot 3.2, Table 1). This model assumed host paths via a gen- Angiosperm eralist state and prohibited reversal from brown rot to specialist 0.03 white rot, which is consistent with our expectation that 0.01 swtiches between angiosperm and gymnosperm Gymnosperm specialization cannot occur in a single step, and that specialist 17.04 losses of AA2s and other lignocellulolytic enzymes 11.01 Generalist makes reversals from brown rot to white rot unlikely. The version of this model that specified the root state with equal weights for white rot states and zero prob- 100 ability for brown rot states (model 3.3) performed better than the model which assumed equal weights among all six states. We display transition rates based on model 65.50 3.3 (Fig. 5). 3.99 We found disparity in rates of transitions between generalism and angiosperm specialization between the 2.04 decay modes. While white rot lineages display high transition rates from generalism to angiosperm Fig. 5 Dynamics of host specialization evolution in wood decay fungi specialization, brown rot lineages display higher rates within Agaricomycetes based on a multi-state likelihood model. from gymnosperm specialization to generalism (Fig. 5). Transition rates based on the best model (“Correlated hosts – norev”, White rot lineages further show higher rates of transi- Table 1) and the maximum likelihood phylogeny among six character states: white or brown rot generalist; white or brown rot angiosperm tions towards angiosperm specialization than the re- specialist and white or brown rot gymnosperm specialist. The six-state verse, whereas brown rot lineages show the opposite, character coding was based on the gymnosperm association. A with higher rates from angiosperm specialization to gen- gymnosperm association above 90% was classified as gymnosperm eralism than the reverse. White and brown rot lineages specialist, below 10% as angiosperm specialist and others as generalists “ – ” both switch more frequently from gymnosperm ( 90 10 specialization ,fordetailsseeTrait data and character matrix). Numbers above and below arrows denote transition rates and the arrow specialization to generalism than the reverse (Fig. 5). width reflects the rate size. For rate estimates based on alternative trees The transition rate estimates were consistent across 100 and the one-genus-subsets see Additional file 1: Figure S5. Angiosperm alternative trees (Additional file 1: Figure S5 A, B). The tree image by Michele M. Tobias (see Acknowledgements) 100 one-genus-subsets yielded consistent relative rates, but rates of white rot states were higher (especially rates from generalists to gymnosperm specialists, Additional file 1: clear picture. The rate estimates based on the ML phyl- Figure S5 A, C). ogeny showed one transition rate from white to brown Concerning the rates of transitions from white to rot angiosperm specialists (Fig. 5). The 100 alternative brown rot estimated based on the ML phylogeny, the al- trees further displayed equally high rates from white to ternative trees and one-genus-subsets did not yield a brown rot generalists (Additional file 1: Figure S5 A, B).

Table 1 The fit of three alternative models of host evolution among decay fungi of Agaricomycetes. The best model (shown in bold), based on Akaike weights (w), was the model 3.3, which allowed only intermediate host transitions (“Correlated hosts”), no brown rot to white rot reversals (“norev”) and a root prior with equal probabilities among white rot fungi and zero probability for brown rot states (“white rot equal”). For model selection based on the exclusivity coding, see Additional file 1: Table S6 Model -Ln L AIC Δ AIC w Uncorrelated, ER − 1870.83 3743.65 1355.72 0.00 Uncorrelated, ARD − 1180.64 2421.27 33.34 0.00 Correlated hosts, ER − 1774.09 3550.19 1162.25 0.00 Correlated hosts, ARD − 1183.56 2395.12 7.19 0.02 Correlated hosts – norev, ER − 1941.71 3885.41 1497.48 0.00 Correlated hosts – norev, ARD, root = equal −1183.66 2389.32 1.39 0.33 Correlated hosts – norev, ARD, root = white rot equal − 1182.97 2387.93 0.00 0.65 Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 10 of 13

Within the 100 alternative trees, brown rot clades were inferred a correlation between brown rot and exclusive not collapsed since SH-like support values were > 90. decay of conifer hosts and suggested that brown rot pro- Transition rates from white to brown rot gymnosperm motes gymnosperm specialization. However, within specialists were either estimated as zero or very low brown rot, transition rates between gymnosperm (Fig. 5, Additional file 1: Figure S5). specialization and generalism are high in both directions, with a trend toward generalism, suggesting that speciali- Discussion zations towards conifer hosts are not stable (Fig. 5). Our Brown rot fungi as a whole comprise a larger proportion findings are robust against topological and branch of gymnosperm specialists than white rot (Fig. 4d), lengths variation (Additional file 1: Figure S5 A, B). Fur- which is consistent with Gilbertson’s observations [31]. ther, our results are robust against different assumptions Nevertheless, most brown rot fungi are generalists and concerning reversals from white to brown rot. Transi- only two of five brown rot clades display mainly gymno- tion rate estimates of the model allowing reversals and sperm specialists (clades Gloeophyllum-Neolentinus and the one disallowing reversals were nearly identical (data Serpula-Hygrophoropsis, Fig. 4d, e). Brown rot lineages not shown, however, AIC difference only 7.19 which is show a higher rate of switches to gymnosperm often considered as not substantially different [78]). specialization than white rot fungi, but brown rot display Further, we estimated the likelihood model of host the highest rate towards generalism. Brown rot further specialization evolution based on the exclusivity coding displayed dynamic transitions between generalism and and found that the transition rate towards gymnosperm specialization (Fig. 5). White rot fungi are highly special- exclusivity was higher for brown rot compared with ized on angiosperm hosts (Figs. 4 and 5). white rot lineages (Additional file 1: Figure S6). This Gilbertson [p. 33] suggested that “85% of brown-rot finding is consistent with the 90–10 specialization cod- polypores occur primarily on conifers”, which was the ing (Fig. 5). Within the exclusivity model we found over- basis for later hypotheses about brown rot evolution in all higher rates from host exclusivity to generalism general [31]. Our analysis could not confirm that brown (Fig. 5, Additional file 1: Figure S6). However, the strin- rot Polyporales occur primarily on gymnosperm hosts gency of this coding scheme may overestimate the num- (Fig. 4e). We found two brown rot clades within Poly- ber of generalist taxa, as species found at extremely high porales, of which the Auriporia-Crustoderma clade con- rates on a single host species (e.g. > 90%, but less than sists of mainly generalists and angiosperm specialists 100%) are still coded as generalists. Thus, rates towards and the Antrodia-Fomitopsis clade of mainly generalists “generalists” are probably overestimated in this coding (Fig. 4e). Thus, neither of the two brown rot clades scheme. Therefore, interpretations from the exclusivity within the Polyporales were mainly specialized on gym- model should be made with caution. For a more detailed nosperms (Fig. 4e). Our dataset allowed us to extend picture, further analysis should thus include three states and evaluate Gilbertson’s statement for a broad range of of host association, separating generalism, non-exclusive brown rot lineages of different clades and orders. Ac- specialization and exclusivity and treat non-exclusive cording to our analysis, only two of five brown rot clades specialization as an intermediate state. consist of mainly gymnosperm specialists, the Gloeophyl- Based on our time-calibrated mega-phylogeny ap- lum-Neolentinus (Gloeophyllales) and the Serpula-Hy- proach, we found that most lineages within Agaricomy- grophoropsis (Boletales) clades (Fig. 4e). Further, the cetes radiated after the origins of gymnosperms and majority of brown rot fungi are generalists (Fig. 4d). angiosperms (Fig. 2). Our estimates for branching times Therefore, the hypothesis that brown rot fungi occur are highly consistent with chronograms of previous stud- primarily on gymnosperms is not generally supported. ies with more limited species sampling, but more gen- Based on our 90–10 specialization coding and a omic information. Floudas et al. [24] for example found multi-state likelihood model of host evolution, we found a mean age of 290 million years for the crown node of that white rot fungi switched frequently between gener- Agaricomycetes, which is consistent with our estimate of alism and angiosperm specialism with a higher rate to- 282 million years (Additional file 1: Figure S7). Smith et wards angiosperm specialism (Fig. 5). Within brown rot al. [79] used an uncorrelated relaxed molecular clock lineages, this pattern shifted towards frequent switches analysis to date a comprehensive plant tree of life and between generalism and gymnosperm specialization found mean crown origins of 301 million years for (Fig. 5). This suggests that brown rot evolution pro- gymnosperms and 217 million years for angiosperms, moted frequent shifts to gymnosperm specialization. respectively. Many of the large clades within Agaricomy- However, the reversal rate from gymnosperm specialism cetes originated before, but diversified after the angio- to generalism is higher, suggesting that specializations sperm and gymnosperm origins (Fig. 2). The estimated towards conifer hosts are not restrictive (Fig. 5). Hibbett timing of origin of the fungal and plant groups is con- and Donoghue [32], based on a much smaller dataset, sistent with our inference that transitions from white rot Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 11 of 13

to brown rot occurred among angiosperm specialists gymnosperm hosts between white and brown rot species. Table S6. The (Fig. 5) or possibly generalists (Additional file 1: Figure fit of three alternative models of host association evolution in white and S5). Relative transition rates in white rot fungi suggest a brown rot lineages of Agaricomycetes based on the exclusivity coding (100–0). (DOCX 22788 kb) pattern of transition away from gymnosperm specialization Additional file 2: The alignment used in maximum likelihood analyses. and towards generalism, followed by relatively higher rates (FAS 43500 kb) of angiosperm specialization (Fig. 5). This pattern away Additional file 3: Newick file for the maximum likelihood phylogeny of from gymnosperm specialization and towards angiosperm the decay fungi of this study. (TRE 56.6 kb) specialization among white rot is consistent with the rela- Additional file 4: Newick file for the comprehensive guide tree used for tively high percentage of white rot angiosperm specialists the maximum likelihood phylogeny. (TRE 27.5 kb) we observed (Fig. 4). Thus, it is plausible that the radiation Additional file 5: Partition scheme for the alignment. The best partition scheme was determined using IQ-Tree, and is listed here. (TXT 153 bytes) of angiosperms created new niches for wood decayers and Additional file 6: Nexus file for the 100 alternative phylogenetic trees of promoted diversification of white rot fungi. the decay fungi of this study. (NEXUS 36100 kb) Additional file 7: GUIDANCE column reliability score for the maximum Conclusion likelihood phylogeny. (TXT 88.5 kb) Our models of host evolution suggest that angiosperms Additional file 8: Data sheet with host associations and decay mode. may have served as a new mega-niche, which was (XLSX 113 kb) exploited particularly well by white rot fungi leading to high specialization rates. Brown rot lineages switched Acknowledgements more frequently towards generalism, suggesting that We thank Heinrich Holzer for permission to use his photograph of Fomes fomentarius. Tree silhouette images in Figs. 4 and 5 were taken from http:// brown rot fungi were limited in exploiting angiosperm phylopic.org. The angiosperm image was made by Michele M. Tobias resources. Whether this limitation on the part of brown (creative commons, https://creativecommons.org/licenses/by-nc-sa/3.0/), and rot in exploiting angiosperm resources is directly related the colour was changed. to the loss in copy number of decay-related genes [26] Funding seems plausible, but remains to be tested by future stud- This research was supported by United States National Science Foundation ies. Moreover, host shifts may be identifiable at the awards IOS-1456777 to DSH. FK was funded by a scholarship of the “Rudolf enzymatic level, if expression patterns for genes coding und Helene Glaser-Stiftung”. Further, this work was supported by the German Research Foundation (DFG) and the Technical University of Munich within for key decay enzymes differ between clades with differ- the funding programme Open Access Publishing. ent host specializations. Such studies represent exciting future possibilities in this system, and may elucidate the Availability of data and materials underlying molecular mechanisms controlling decay Availability of data We have made the nucleotide alignment, the maximum likelihood mode shifts. phylogenies (and 100 alternative trees), the GUIDANCE column score, the partition file and the guide tree used in this study, available with online supplemental material as Additional files 2, 3, 4, 5, 6, 7 and 8. Additional file Software Project name: R package rusda Project home page: https://github.com/ropensci/rusda, https://cran.r- Additional file 1: rusda: an R interface to the United States Department project.org/web/packages/rusda/index.html of Agriculture’s Fungus-Host Distribution Database. Figure S1. Genomic Programming language: R phylogenetic tree compiled from Floudas et al. 2012; Kohler et al. 2015; License: GNU GPL Nagy et al. 2015 used as the backbone for the comprehensive guide tree for the RAxML tree inference. Figure S2. Maximum Likelihood phylogeny Authors’ contributions of the Agaricomycetes with color coding for 14 orders. Figure S3. FK and DSH designed the study and drafted the manuscript; FK carried out Phylogeny with SH support values. Figure S4. Transition rates between the statistical analyses; FK, CB and HS drafted earlier versions of the the states in 6 × 6 Q-matrices with six states: (1) white rot/angiosperm manuscript. JS helped with statistical method selection and drafted the specialist, (2) brown rot/angiosperm specialist, (3) white rot/gymnosperm manuscript. FK and CH carried out sequence assembly. All authors read and specialist, (4) brown rot/ gymnosperm specialist, (5) white rot/generalist, commented on the final version of the manuscript. and (6) brown rot/generalist. Figure S5. Transition rates among six character states based on a maximum likelihood (ML) phylogeny, 100 alternative trees and the one-genus-subset (100 times bootstrapped). Ethics approval and consent to participate Figure S6. Dynamics of host specialization in wood decay fungi within Not applicable. the Agaricomycetes based on the 100–0 exclusivity coding. Figure S7. Branching times for backbone of time-dated Agaricomycetes phylogeny. Consent for publication “ ” Figure S8 Overview of query results using R package rusda based on Not applicable. the Fungus-Host Distribution Database (FHDD) using 29,591 Dikarya and 105,350 Spermatophyta species as input. Table S1. Best partition scheme Competing interests found by IQ-Tree [11, 12]. Table S2. Tip state frequencies of white and The authors declare that they have no competing interests. brown rot specialization based on different thresholds of host association [%]. Table S3. Re-classification of taxonomic orders based on Binder et al., 2005; Hibbett et al., 2014; Larsson, Larsson, & Kõljalg, 2004. Table S4. ’ Table of reference species and association numbers from NCBI. Table S5 Publisher sNote Phylogenetic and normal linear regression on the number of angio- and Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 12 of 13

Author details 25. Nagy LG, Ohm RA, Kovács GM, Floudas D, Riley R, Gácser A, et al. Latent 1Plant Biodiversity Research Group, Center for Food and Life Sciences homology and convergent regulatory evolution underlies the repeated Weihenstephan, Technische Universität München, Freising, Germany. emergence of yeasts. Nat Commun. 2014;5 2Baverian Forest National Park, Grafenau, Germany. 3Department of 26. Kohler A, Kuo A, Nagy LG, Morin E, Barry KW, Buscot F, et al. Convergent Environmental Science, The Connecticut Agricultural Experiment Station, losses of decay mechanisms and rapid turnover of symbiosis genes in New Haven, CT 06511, USA. 4Biology Department, Clark University, Worcester, mycorrhizal mutualists. Nat Genet. 2015;47:410–5. MA 01610, USA. 27. Lombard V, Golaconda Ramulu H, Drula E, Coutinho PM, Henrissat B. The carbohydrate-active enzymes database (CAZy) in 2013. Nucleic Acids Res. Received: 17 November 2017 Accepted: 3 July 2018 2014;42:D490–5. 28. Hofrichter M, Ullrich R, Pecyna MJ, Liers C, Lundell T. New and classic families of secreted fungal heme peroxidases. Appl Microbiol Biotechnol. 2010:871–97. 29. Nagy LG, Riley R, Bergmann, PJ Krizsán, K, Martin FM, Grigoriev I V, Cullen D, References Hibbett DS. Genetic bases of fungal white rot wood decay predicted by 1. McCarl BA, Metting FB, Rice C. Soil carbon sequestration. Clim Change. phylogenomic analysis of correlated gene-phenotype evolution. Mol Biol 2007;80:1–3. Evol. 2016;34:35–44. 2. Siegenthaler U, Sarmiento JL. Atmospheric carbon dioxide and the ocean. 30. Burdsall HH, Blackwell M, Nakasone KK. Robert Lee Gilbertson, 1925–2011. Nature. 1993;365:119–25. Mycologia Mycol Soc Am. 2012;104:1521–4. 3. Horwath W. 12 – carbon cycling and formation of soil organic matter. Soil 31. Gilbertson RL. Wood-rotting fungi of North America. Mycologia. 1980;72:1–49. Microbiol Ecol Biochem. 2007:303–39. 32. Hibbett DS, Donoghue MJ. Analysis of character correlations among wood 4. Welker C, Balasubramanian V, Petti C, Rai K, DeBolt S, Mendu V. Engineering decay mechanisms, mating systems, and substrate ranges in plant biomass lignin content and composition for biofuels and bioproducts. homobasidiomycetes. Syst Biol. 2001;50:215–42. Energies. 2015;8:7654–76. 33. Farr DF, Rossman AY, Palm ME, McCray EB. Fungal databases, systematic 5. Lundell TK, Mäkelä MR, Hildén K. Lignin-modifying enzymes in filamentous mycology and microbiology laboratory [Internet]. ARS, USDA. 2012. basidiomycetes–ecological, functional and phylogenetic review. J Basic Available from: http://nt.ars-grin.gov/fungaldatabases/ Microbiol. 2010;50:5–20. 34. Tedersoo L, Bahram M, Polme S, Koljalg U, Yorou NS, Wijesundera R, et al. 6. Medie FM, Davies GJ, Drancourt M, Henrissat B. Genome analyses Global diversity and geography of soil fungi. Science. 2014;346:6213. highlight the different biological roles of cellulases. Nat Rev Microbiol. 35. Kim SY, Park SY, Ko KS, Jung HS. Phylogenetic analysis of Antrodia and 2012;10:227–34. related taxa based on partial mitochondrial SSU rDNA sequences. Antonie 7. Sjöstrom E, Wood Chemistry: Fundamentals and Applications, Academic Van Leeuwenhoek. 2003;83:81–8. Press, New York, 1981. 36. Molina FI, Shen P, Jong SC, Orikono K. Molecular evidence supports the 8. Bugg TDH, Ahmad M, Hardiman EM, Rahmanpour R. Pathways for separation of Lentinula edodes from Lentinus and related genera. Can J Bot degradation of lignin in bacteria and fungi. Nat Prod Rep. 2011;28:1883–96. Can Bot. 1992;70:2446–52. 9. Heimann M, Reichstein M. Terrestrial ecosystem carbon dynamics and 37. Rathod MM. Taxonomic studies on the Daedaloid and Hexagonoid climate feedbacks. Nature. 2008;451:289–92. Polypores form the Forest of western Maharasta. Recent Res Sci Technol. 10. Harley JL. Fungi in ecosystems. J Appl Ecol JSTOR. 1971:627–42. 2011;3 11. Beech E, Rivers M, Oldfield S, Smith PP. GlobalTreeSearch: the first complete 38. Barrasa JM, Esteve-Raventós F, Dähncke RM. Clitocybula canariensis global database of tree species and country distributions. J Sustain For. (Tricholomataceae), a new brown-rot fungus from the Canary Islands 2017;36:454–89. (Spain). Fungal Divers. 2006;22:1–11. 12. Thakur VK, Thakur MK. Processing and characterization of natural cellulose 39. Wesenberg D, Buchon F, Agathos SN. Degradation of dye-containing textile fibers/thermoset polymer composites. Carbohydr Polym. 2014:102–17. effluent by the agaric white-rot fungus Clitocybula dusenii. Biotechnol Lett. 13. Cornwell WK, Cornelissen JHC, Allison SD, Bauhus J, Eggleton P, Preston CM, 2002;24:989–93. et al. Plant traits and wood fates across the globe: rotted, burned, or 40. Gilbertson RL, Martin KJ, Lindsey JP. Annotated check list and host index for consumed? Glob Chang Biol. 2009;15:2431–49. Arizona wood-rotting fungi. College of Agriculture. Tucson: University of 14. Pichersky E, Gang DR. Genetics and biochemistry of secondary metabolites Arizona; 1974. in plants: an evolutionary perspective. Trends Plant Sci. 2000:439–45. 41. Petersen RH. Gloeomucro and a note on Physalacria concinna. Mycologia. 15. Pallardy SG. Physiology of Woody plants. Physiol Woody Plants. 2008; 1980:301–11. 16. Wagenführ R, Scheiber C. Holzatlas. Fachbuchverlag Leipzig; 2007. 42. Heibl C. The megaptera package: Large phylogenetic dataset assembly in R. 17. Worrall JJ, Anagnost SE, Zabel RA. Comparison of wood decay among R package version 1.0–25. 2014. diverse lignicolous fungi. Mycologia JSTOR. 1997:199–219. 43. FitzJohn RG, Pennel MW, Zanne AE, Stevens PF, Tank DC, Cornwell WK. How 18. Blanchette RA. Delignification by wood-decay fungi. Annu. Rev. much of the world is woody? J Ecol. 2014;102:1266–72. Phytopathol. 1991;29:381–403. 44. Heibl C. The megapera package: Large phylogenetic dataset assembly in R. 19. Lundell TK, Mäkelä MR, de Vries RP, Hildén KS. Genomics, lifestyles and Modern Phylogenetic Comparative Methods and their application in future prospects of wood-decay and litter-decomposing basidiomycota. evolutionary biology. Seville, Spain,. 11–15.11.2014. 2014. Adv Bot Res. 2014;70:329–70. 45. Smith SA, Beaulieu JM, Donoghue MJ. Mega-phylogeny approach for 20. Schneider T, Keiblinger KM, Schmid E, Sterflinger-Gleixner K, Ellersdorfer G, comparative biology: an alternative to supertree and supermatrix Roschitzki B, et al. Who is who in litter decomposition? Metaproteomics approaches. BMC Evol Biol. 2009;9:37. reveals major microbial players and their biogeochemical functions. ISME J. 46. Benson DA, Cavanaugh M, Clark K, Karsch-Mizrachi I, Lipman DJ, Ostell J, 2012;6:1749–62. Sayers EW. GenBank. Nucleic Acids Res. 2012;41:D36–D42. 21. Riley R, Salamov AA, Brown DW, Nagy LG, Floudas D, Held BW, et al. 47. Binder M, Justo A, Riley R, Salamov A, Lopez-Giraldez F, Sjokvist E, et al. Extensive sampling of basidiomycete genomes demonstrates inadequacy of Phylogenetic and phylogenomic overview of the Polyporales. Mycologia. the white-rot/brown-rot paradigm for wood decay fungi. Proc Natl Acad Sci 2013;105:1350–73. U S A. 2014;111:9923–8. 48. Sela I, Ashkenazy H, Katoh K, Pupko T. GUIDANCE2: accurate detection of 22. Rineau F, Shah F, Smits MM, Persson P, Johansson T, Carleer R, et al. Carbon unreliable alignment regions accounting for the uncertainty of multiple availability triggers the decomposition of plant litter and assimilation of parameters. Nucleic Acids Res. 2015;43:W7–W14. nitrogen by an ectomycorrhizal fungus. ISME J. 2013;7:2010–22. 49. Penn O, Privman E, Landan G, Graur D, Pupko T. An alignment confidence 23. Maijala P, Fagerstedt KV, Raudaskoski M. Detection of extracellular score capturing robustness to guide tree uncertainty. Mol Biol Evol. 2010;27: cellulolytic and proteolytic activity in ectomycorrhizal fungi and 1759–67. Heterobasidion annosum (Fr.) Bref. New Phytol. 1991;117:643–8. 50. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: 24. Floudas D, Binder M, Riley R, Barry K, Blanchette RA, Henrissat B, et al. The improvements in performance and usability. Mol Biol Evol. 2013;30:772–80. Paleozoic origin of enzymatic lignin decomposition reconstructed from 31 51. Landan G, Graur D. Local reliability measures from sets of co-optimal fungal genomes. Science. 2012;336:1715–9. multiple sequence alignments. Pacific Symp Biocomput. 2008;13:15–24. Krah et al. BMC Evolutionary Biology (2018) 18:119 Page 13 of 13

52. Tan G, Muffato M, Ledergerber C, Herrero J, Goldman N, Gil M, et al. Current methods for automated filtering of multiple sequence alignments frequently worsen single-gene phylogenetic inference. Syst Biol. 2015;64:778–91. 53. Nguyen LT, Schmidt HA, Von Haeseler A, Minh BQ. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol. 2015;32:268–74. 54. Chernomor O, Von Haeseler A, Minh BQ. Terrace aware data structure for Phylogenomic inference from Supermatrices. Syst Biol. 2016;65:997–1008. 55. Stamatakis A, Hoover P, Rougemont J. A rapid bootstrap algorithm for the RAxML web servers. Syst. Biol. 2008;57:758–71. 56. Miller MA, Pfeiffer W, Schwartz T. Creating the CIPRES science gateway for inference of large phylogenetic trees. Gatew Comput Environ Work. 2010; 2010:1–8. 57. Miller MA, Pfeiffer W, Schwartz T. The CIPRES science gateway: a community resource for phylogenetic analyses. Proc. 2011 TeraGrid Conf. Extrem. Digit. Discov. 2011. p. 41. 58. Anisimova M, Gil M, Dufayard J-F, Dessimoz C, Gascuel O. Survey of branch support methods demonstrates accuracy, power, and robustness of fast likelihood-based approximation schemes. Syst. Biol. 2011;60:685–99. 59. Paradis E, Claude J, Strimmer K. APE: Analyses of phylogenetics and evolution in R language. Bioinformatics. 2004;20:289–90. 60. Hibbett D, Grimaldi D, Donoghue M. Fossil mushrooms from Miocene and cretaceous ambers and the evolution of Homobasidiomycetes. Am. J. Bot. 1997;84:981. 61. LePage BA, Currah RS, Stockey AR, Rothwell GW. Fossil ectomycorrhizae from the middle Eocene. Am J Bot. 1997;84:410. 62. Paradis E. Molecular dating of phylogenies by likelihood methods: a comparison of models and a new information criterion. Mol Phylogenet Evol. 2013;67:436–44. 63. Drummond AJ, Bouckaert RR. Bayesian evolutionary analysis with BEAST 2. Bayesian Evol. Anal. with BEAST. 2015;249. 64. Aberer AJ, Kobert K, ExaBayes SA. Massively parallel Bayesian tree inference for the whole-genome era. Mol Biol Evol. 2014;31:2553–6. 65. Drummond AJ, Suchard MA, Xie D, Rambaut A. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol. Biol. Evol. 2012;29:1969–73. 66. Ho LST, Ané C. A linear-time algorithm for Gaussian and non-Gaussian trait evolution models. Syst Biol. 2014;63:397–408. 67. Pagel M. Inferring the historical patterns of biological evolution. Nature. 1999;401:877–84. 68. Beaulieu JM, Oliver JC, O’Meara B. Package `corHMM’. 2014. 69. Akaike H. A new look at the statistical model identification. IEEE Trans Autom Control. 1974;19:716–23. 70. Nagy LG, Floudas D, Riley R, Barry K, Grigoriev IV, Hibbett DS. Diversification of wood decay systems in early evolution of Agaricomycotina. Phytopathology. 2013:181. 71. Nagy LG, Riley R, Tritt A, Adam C, Daum C, Floudas D, et. al. Comparative Genomics of Early-Diverging Mushroom-Forming Fungi Provides Insights into the Origins of Lignocellulose Decay Capabilities. Mol. Biol. Evol. 2015;33:959–70. 72. Beaulieu JM, Donoghue MJ. Fruit evolution and diversification in Campanulid angiosperms. Evolution. 2013;67:3132–44. 73. Fritz SA, Purvis A. Selectivity in mammalian extinction risk and threat types: a new measure of phylogenetic signal strength in binary traits. Conserv Biol. 2010;24:1042–51. 74. Orme D. The caper package : comparative analysis of phylogenetics and evolution in R. R Packag version 0.5, 2. 2013;1–36. 75. Revell LJ. phytools: An R package for phylogenetic comparative biology (and other things). Methods Ecol. Evol. 2012;3:217–23. 76. Keck F, Rimet F, Bouchez A, Franc A. Phylosignal: an R package to measure, test, and explore the phylogenetic signal. Ecol. Evol. 2016;6:2774–80. 77. Bush SE, Weckstein JD, Gustafsson DR, Allen J, DiBlasi E, Shreve SM, et al. Unlocking the black box of feather louse diversity: a molecular phylogeny of the hyper-diverse genus Brueelia. Mol. Phylogenet. Evol. 2016;94:737–51. 78. Burnham KP, Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach. Springer Science & Business Media, 2003. 79. Smith SA, Beaulieu JM, Donoghue MJ. An uncorrelated relaxed-clock analysis suggests an earlier origin for flowering plants. Proc Natl Acad Sci. 2010;107:5897–902.